You will work across the full model lifecycle - from dataset creation and experimentation through training, evaluation, deployment, monitoring, and iteration.
This role is a full-time position located out of our office in Arlington, VA / Pittsburgh, PA / San Francisco, CA / Anywhere in the US.
This role may require up to 25% travel, including periodic travel to our Pittsburgh, PA and Arlington, VA offices for team collaboration, planning activities, and in-person meetings.
Scope of Responsibilities
Design and build LLMOps infrastructure supporting the development, evaluation, deployment, and continuous improvement of production language models.
Build scalable training and fine-tuning infrastructure for commercial and open-weight language models.
Develop pipelines supporting supervised fine-tuning, parameter-efficient fine-tuning, preference optimization, and other post-training techniques.
Build infrastructure for distributed training and GPU-accelerated ML workloads.
Develop data pipelines for training, fine-tuning, evaluation, and synthetic data generation.
Build systems for dataset versioning, lineage, quality validation, transformation, and reproducible experimentation.
Develop experiment management infrastructure that enables engineers to compare models, datasets, hyperparameters, prompts, and training techniques.
Build automated evaluation pipelines that determine whether new models or model versions are ready for production deployment.
Design model registries, artifact management, versioning, and promotion workflows across development and production environments.
Build and operate scalable model-serving and inference infrastructure for open-weight and fine-tuned models.
Develop abstractions that allow product and AI engineering teams to use multiple models and inference providers without tightly coupling applications to a single model or vendor.
Build observability for model training and inference, including metrics, tracing, logging, resource utilization, model quality, latency, throughput, and cost.
Optimize training and inference workloads for GPU utilization, throughput, latency, reliability, and infrastructure cost.
Build automated workflows for model deployment, rollback, canarying, and production validation.
Investigate model and infrastructure failures across data pipelines, training jobs, inference services, distributed systems, and production environments.
Evaluate emerging models, training techniques, inference frameworks, and ML infrastructure and determine where they can improve our production systems.
Partner closely with AI engineers building agentic systems to provide the model, evaluation, and training infrastructure required to continuously improve those systems.
Qualifications
U.S. Citizenship is required
Required Skills
5+ years of experience building production machine learning systems, ML infrastructure, distributed systems, or similar technical systems.
Deep experience designing, building, and operating production ML infrastructure or ML platforms.
Experience building infrastructure for training, fine-tuning, evaluating, deploying, and monitoring large language models or other large-scale deep learning models.
Experience with LLM fine-tuning and post-training workflows, including techniques such as supervised fine-tuning, LoRA/QLoRA or other parameter-efficient approaches, and preference optimization.
Strong understanding of the modern LLM lifecycle, including data preparation, training, evaluation, model artifacts, deployment, inference, monitoring, and iteration.
Experience building reproducible ML pipelines involving dataset versioning, experiment tracking, model versioning, and automated evaluation.
Experience building and operating production GPU infrastructure across AWS, GCP, Azure, or dedicated GPU providers, including training and/or inference workloads.
Strong understanding of distributed systems and the challenges involved in running computationally intensive ML workloads at scale.
Strong programming experience in Python and experience building production-quality software.
Deep experience with containers, Kubernetes, and cloud platforms such as AWS, GCP, or Azure.
Experience designing scalable APIs, services, asynchronous workloads, and data-processing pipelines.
Strong understanding of observability and operational reliability for production ML systems.
Comfortable debugging failures across training code, datasets, models, GPUs, distributed systems, and cloud infrastructure.
Able to move between ML experimentation and infrastructure engineering, understanding the needs of researchers and AI engineers while building systems that make those workflows scalable and reproducible.
Comfortable working in a rapidly evolving field where tooling, models, and best practices change quickly.
Desired Skills
Current possession of a U.S. security clearance, or the ability to obtain one with our sponsorship
Experience in or exposure to the nuances of a startup or other entrepreneurial environment
Experience building secure code execution environments or sandboxes for AI agents.
Experience with multi-agent architectures, agent-to-agent communication, or distributed agent execution.
Experience with fine-tuning, post-training, reinforcement learning, or synthetic data generation.
Experience building AI observability, tracing, and debugging infrastructure.
Experience optimizing inference latency, throughput, GPU utilization, or model-serving costs.
Experience with AI security, adversarial testing, or securing agentic systems.
Experience working in government, defense, or other mission-critical environments.
We firmly believe that past performance is the best indicator of future performance. If you thrive while building solutions to complex problems, are a self-starter, and are passionate about making an impact in global security, we’re eager to hear from you.
Air is an Equal Opportunity Employer. All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, national origin, sexual orientation, gender identity, disability and protected veterans status or any other characteristic protected by law.
